Science Question Answering on ScienceQA Image (Score)
95.6ScoreLLaVA-OV-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-OV-7BKV Size=1962026.05 | 95.6 | |
| LLaVA-OV-7B + HERMESKV Size=1002026.05 | 90.9 | |
| LLaVA-OV-7B + HERMESKV Size=402026.05 | 87.6 | |
| Cambrian-4B + Gaze attentionKV Size=72+42026.05 | 82.1 | |
| Cambrian-4B + Gaze attentionKV Size=144+42026.05 | 82 | |
| Cambrian-4B + Gaze attentionKV Size=288+42026.05 | 81.5 | |
| Cambrian-4B + TokenCompre.KV Size=1442026.05 | 80.7 | |
| Qwen2.5-VL-3BKV Size=2562026.05 | 80.5 | |
| Qwen2.5-VL-3B + InfiniPot-VKV Size=1282026.05 | 80.5 | |
| Cambrian-4BKV Size=5762026.05 | 79.9 | |
| Cambrian-4B + TokenCompre. + Gaze attentionKV Size=36+42026.05 | 79.5 | |
| Cambrian-4B + TokenCompre. + Gaze attentionKV Size=54+42026.05 | 79.3 | |
| Qwen2.5-VL-3B + InfiniPot-VKV Size=502026.05 | 79 | |
| Cambrian-8B-CLIP-LKV Size=5762026.05 | 77.4 | |
| FastVBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 74.2 | |
| VisionZipBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 74.2 | |
| HoloVBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 74.2 | |
| VisPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 74 | |
| VisPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 73.9 | |
| VisPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 73.9 | |
| VisionZipBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 73.8 | |
| FSRBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 73.8 | |
| HoloVBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 73.6 | |
| LLaVA-Next-Qwen2-7BKV Size=7842026.05 | 73.5 | |
| FSRBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 73.3 | |
| LLaVA-NeXT-13BBackbone=LLaVA-NeXT-13B, Retained Tokens=2880, Pruning Ratio=0%2026.02 | 73.2 | |
| FastVBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 73.1 | |
| FSRBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 73 | |
| CoVFTBackbone=LLaVA-1.5-13B2026.03 | 72.9 | |
| LLaVA-Next-LLaMA-3-8BKV Size=7842026.05 | 72.9 | |
| LLaVA-1.5-13BBackbone=LLaVA-1.5-13B, Retained Tokens=576, Pruning Ratio=0%2026.02 | 72.8 | |
| DivPruneBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 72.8 | |
| CDPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 72.5 | |
| FreezeBackbone=LLaVA-1.5-13B2026.03 | 72.43 | |
| CDPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 72.4 | |
| CDPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 72.4 | |
| HoloVBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 72.3 | |
| DivPruneBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 72.3 | |
| FSRBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 72.2 | |
| DivPruneBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 72.2 | |
| BitFitBackbone=LLaVA-1.5-13B2026.03 | 72.14 | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 72.1 | |
| Full fine-tuningBackbone=LLaVA-1.5-13B2026.03 | 71.99 | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 71.9 | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 71.9 | |
| DivPruneBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 71.7 | |
| FastVBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 71.7 | |
| HoloVBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 71.7 | |
| FSRBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 71.6 | |
| LLaVA-1.5Backbone=LLaVA-1.5-13B2026.03 | 71.6 | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 71.1 | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 71.1 | |
| VisionZipBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 70.8 | |
| HoloVBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 70.6 | |
| FastVBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 70.5 | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 70.5 | |
| FSRBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 70.3 | |
| VisionZipBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 70.2 | |
| CoVFTBackbone=LLaVA-1.5-7B2026.03 | 69.51 | |
| HoloVBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 69.5 | |
| FreezeBackbone=LLaVA-1.5-7B2026.03 | 69.31 | |
| BitFitBackbone=LLaVA-1.5-7B2026.03 | 68.86 | |
| SVPTBackbone=LLaVA-1.5-7B2026.03 | 68.82 | |
| Cambrian-8B-DINOv2-LKV Size=5762026.05 | 68.7 | |
| SPHINX-7B2026.05 | 68.6 | |
| VPTBackbone=LLaVA-1.5-7B2026.03 | 67.92 | |
| Full fine-tuningBackbone=LLaVA-1.5-7B2026.03 | 67.67 | |
| LoRABackbone=LLaVA-1.5-7B2026.03 | 67.63 | |
| LLaVA-1.5Backbone=LLaVA-1.5-7B2026.03 | 66.8 | |
| BLIP-2-7BKV Size=5762026.05 | 63 |